Peroxidasin Is Secreted and Incorporated into the Extracellular Matrix of Myofibroblasts and Fibrotic Kidney

Matrix Pathobiology

Peroxidasin Is Secreted and Incorporated into theExtracellular Matrix of Myofibroblasts and FibroticKidney

Zalan Peterfi,* Agnes Donko,* Anna Orient,*Adrienn Sum,* Agnes Prokai,† Beata Molnar,*Zoltan Vereb,‡ Eva Rajnavolgyi,‡

Krisztina J. Kovacs,§ Veronika Muller,¶

Attila J. Szabo,† and Miklos Geiszt*From the Department of Physiology,* the First Department of

Pediatrics,† and the Department of Pulmonology,¶ Semmelweis

University, Faculty of Medicine, Budapest; the Institute of

Immunology,‡ University of Debrecen, Medical and Health

Science Center, Debrecen; and the Laboratory of Molecular

Neuroendocrinology,§ Institute of Experimental Medicine,

Budapest, Hungary

Mammalian peroxidases are heme-containing en-zymes that serve diverse biological roles, such as hostdefense and hormone biosynthesis. A mammalian ho-molog of Drosophila peroxidasin belongs to the per-oxidase family; however, its function is currently un-known. In this study, we show that peroxidasin ispresent in the endoplasmic reticulum of humanprimary pulmonary and dermal fibroblasts , and theexpression of this protein is increased during trans-forming growth factor-�1-induced myofibroblast dif-ferentiation. Myofibroblasts secrete peroxidasin intothe extracellular space where it becomes organizedinto a fibril-like network and colocalizes with fi-bronectin, thus helping to form the extracellular ma-trix. We also demonstrate that peroxidasin expres-sion is increased in a murine model of kidney fibrosisand that peroxidasin localizes to the peritubularspace in fibrotic kidneys. In addition, we show thatthis novel pathway of extracellular matrix formationis unlikely mediated by the peroxidase activity of theprotein. Our data indicate that peroxidasin secretionrepresents a previously unknown pathway in extra-cellular matrix formation with a potentially impor-tant role in the physiological and pathological fibro-genic response. (Am J Pathol 2009, 175:725–735; DOI:

10.2353/ajpath.2009.080693)

Peroxidases are heme-containing enzymes with highlyconserved structure, serving diverse functions in theplant and animal kingdom.1 Peroxidases catalyze theoxidation of various substrates in the presence of H2O2.Mammalian peroxidases have an important role in sev-eral physiological processes including host defense andhormone biosynthesis. The family of mammalian per-oxidases consists of myeloperoxidase, eosinophil per-oxidase, lactoperoxidase, thyroid peroxidase, and themammalian peroxidasin. Myeloperoxidase, eosinophilperoxidase, and lactoperoxidase have antimicrobial ac-tivity and serve in the first line of host defense, whilethyroid peroxidase has an essential role in the biosynthe-sis of thyroid hormones.2–4 The function of the mamma-lian peroxidasin is currently unknown. Peroxidases inplants and in lower animal species frequently participatein extracellular matrix (ECM) formation. In the presence ofH2O2, peroxidases enzymatically cross-link extracellularproteins through tyrosine residues.5 ECM stabilization bydityrosine bridges is well-documented during sea urchinfertilization, where secreted ovoperoxidase is responsi-ble for the formation of cross-links.6 Dityrosine formationis also involved in the stabilization of C. elegans cuticle,where dual oxidases, carrying both NADPH oxidase andperoxidase-like domains, provide hydrogen peroxide forthe crosslinking reaction.7

Peroxidasin (PXDN), a unique form of peroxidase wasfirst identified in Drosophila melanogaster.8 Beside con-taining a peroxidase domain, which is highly homologousto other animal peroxidases, peroxidasin also containsprotein domains characteristic for proteins of the ECM.Drosophila PXDN was found to be expressed in severalstages of development, but the exact function remained

Supported by grants from the Hungarian Research Fund (OTKA 042573and NF72669) and the Cystic Fibrosis Foundation (USA) and by grantsfrom the Jedlik Anyos program (1/010/2005). Miklos Geiszt is recipient ofa Wellcome Trust International Senior Fellowship.

Z.P. and A.D. contributed equally to this work.

Accepted for publication April 23, 2009.

Address reprint requests to Miklos Geiszt, Department of Physiology,Semmelweis University, Faculty of Medicine, PO Box 259 H-1444 Buda-pest, Hungary. E-mail: [email protected].

The American Journal of Pathology, Vol. 175, No. 2, August 2009

Copyright © American Society for Investigative Pathology

DOI: 10.2353/ajpath.2009.080693

725

unknown.8 Little is still known about the mammalianPXDN protein. A human homolog of Drosophila PXDN wasoriginally identified as a p53-responsive gene productfrom a colon cancer cell line, but it was not characterizedin detail.9 An independent cloning effort, using subtrac-tive hybridization also led to the identification of the mam-malian PXDN gene, which was originally named mela-noma gene 50, based on the expression in melanomasamples.10 This latter study has characterized PXDN as apossible potent melanoma-associated antigen, but it didnot examine the possible physiological role of the protein.

Here we demonstrate that peroxidasin is expressed byhuman primary cells, including fibroblasts of differentorigin, where the protein is localized to the endoplasmicreticulum. On stimulation by transforming growth factor(TGF)-�1, differentiating myofibroblasts show increasedexpression of peroxidasin. The protein becomes secretedto the extracellular space where it is organized into a fibril-like network. We also show that this pathway of ECM forma-tion is probably not mediated by the peroxidase activity ofthe protein. Our results suggest that beside the secre-tion of well-known constituents of the ECM, PXDN se-cretion by myofibroblasts is a novel way of ECM mod-ification in wound repair and tissue fibrosis.

Materials and Methods

Materials

We used the following antibodies in our studies: Alexa488-and Alexa568-labeled anti-rabbit and anti-mouse Fab (Mo-lecular Probes, Eugene, OR), protein disulfide isomerase(PDI) antibody (RL90), and fibronectin antibody (IST-9) (Ab-cam, Cambridge, UK), lamin antibody (Santa Cruz Bio-technology, Santa Cruz, CA), �-actin antibody, andsmooth muscle actin (SMA) antibody (Sigma ChemicalCo., St. Louis, MO).

Anti-PXDN Antibody

PXDN polyclonal antibody was purified from rabbit se-rum following intracutaneous injections of glutathione

S-transferase-PXDN (amino acids 1329 to 1479). Theantibody was affinity purified using Affigel 10 beads(BioRad Laboratories, Richmond, CA) loaded with theantigen.

Cell Culture and Treatments

COS-7 cells were grown in Dulbecco’s Modified EaglesMedium with Glutamax I (Invitrogen Corp., Carlsbad, CA)supplemented with 10% fetal calf serum, 50 U/ml pen-icillin (Sigma), and 50 �g/ml streptomycin (Sigma).Human pulmonary and dermal fibroblasts (PromoCell,Heidelberg, Germany) were grown in fibroblast basalmedium supplemented with 2% fetal calf serum, 5�g/ml insulin, and 1 ng/ml basic fibroblast growth fac-tor. Cells were grown in a humidified atmosphere of 5%CO2 in air at 37°C. Before TGF-�1 treatment, primaryfibroblasts were serum-deprived in the presence of0.05% serum. Cells were treated with TGF-�1 (R&DSystems, Minneapolis, MN) for 24 to 72 hours in theabsence of serum. In some experiments the mediumwas supplemented with 500 �mol/L �-aminolevulonicacid (Sigma).

Transient Transfections

PXDN encoding pcDNA 3.1 plasmid was transfectedby using Fugene6 (Roche Diagnostics GmbH, Mann-heim, Germany) or Lipofectamine2000 (Invitrogen). Smallinterfering (si)RNA was transfected in 100 nmol/L con-centration using the Interferin siRNA transfection re-agent (Polyplus Transfection, Illkirch, France) orRNAiMAX (Invitrogen).

siRNA Sequences

Sequence-specific and control Stealth siRNAs were ob-tained from Invitrogen. The sequences of PXDN-specificand control siRNAs are provided in Table 1.

Table 1. siRNA Sequences

Name Sequence

PXDN-1 5�-CCUCCAUCCUAGAUCUUCGCUUUAA-3�PXDN-1control 5�-CCUCCCUCAUAGAUGUUCCCUUUAA-3�PXDN-2 5�-GCAUAACAACCGGAUUACACAUUUA-3�PXDN2control 5�-GCAUCAAAACCGGAUAACUCAUUUA-3�

Oligonucleotides Used in Quantitative PCR Experiments

Name Species Sequence

gabdh F H. Sapiens 5�-AAGGTGAAGGTCGGAGTCAACGG-3�gabdh R H. Sapiens 5�-CCAAAGTTGTCATGGATGACCTTGG-3�pxdn F H. Sapiens 5�-CTCAGCCTTCAGCACACGCTC-3�pxdn R H. Sapiens 5�-GAGTTCTGGGTGTTTCCTGGT-3�gabdh F M. musculus 5�-CTGAGTATGTCGTGGAGTCTACTG-3�gabdh R M. musculus 5�-AAGGCCATGCCAGTGAGCTTC-3�pxdn F M. musculus 5�-CGAGGCCGGGACCATGGCATC-3�pxdn R M. musculus 5�-CTGCAGGCTGGCAAGCTTCCAC-3�

726 Peterfi et alAJP August 2009, Vol. 175, No. 2

Western Blot Experiments

Cells lysed in Laemmli sample buffer were boiled andrun on 7.5% or 10% polyacrylamide gels. After blottingonto nitrocellulose membranes blocking was per-formed in PBS 5% milk and 0.1% Tween 20 for 1 hourat room temperature. We incubated the membraneswith the first antibody for 1 hour at room temperature.Membranes were washed five times in PBS 0.1%Tween 20 and horseradish peroxidase-labeled anti-rabbit secondary antibody (Amersham Pharmaceuti-cals, Amersham, UK) was used in 1:5000 dilution andsignals were detected on FUJI Super RX films using theenhanced chemiluminescence method. To precipitatePXDN from the cell culture medium, the medium wasremoved, and 1 volume of 100% (w/v) trichloroacetic acidwas added to four volumes of medium. The samples werekept on ice for 10 minutes then the precipitate was sepa-rated by centrifugation (14,000 rpm, 5 minutes). The pelletswere washed three times with 2 ml of cold acetone anddried by placing the tube in 95°C heat block for 5 minutes.The pellets were resuspended in 4� sample buffer thenboiled for 10 minutes before they were loaded onto poly-acrylamide gels.

Measurement of Peroxidase Activity

COS-7 cells expressing PXDN or primary fibroblasts werelysed in PBS containing 1% hexadecyltrimethylammo-nium bromide. Peroxidase activity of the lysates wasimmediately determined by the Amplex Red peroxidaseassay (Molecular Probes). After 30 minutes incubationtime with the Amplex Red reagent, resorufin fluorescencewas measured at 590 nm.

Measurement of H2O2 Production

TGF-�1-induced H2O2 production of pulmonary fibro-blasts was measured with the Amplex Red assay (Mo-lecular Probes). Attached cells were incubated in thepresence of 50 �mol/L Amplex Red and 0.1 U/ml horse-radish peroxidase in an extracellular medium containing145 mmol/L NaCl, 5 mmol/L KCl, 1 mmol/L MgCl2, 0.8mmol/L CaCl2, 5 mmol/L glucose, and 10 mmol/L HEPES.After 1 hour incubation at 37°C, resorufin fluorescencewas measured at 590 nm.

Immunofluorescent Labeling and ConfocalLaser Microscopy

Cells grown on coverslips were fixed in 4% paraformal-dehyde in PBS then rinsed 5 times in PBS and incubatedfor 10 minutes in PBS containing 100 mmol/L glycine.Coverslips were washed twice in PBS and permeabilizedin PBS containing 1% bovine serum albumin and 0.1%Triton X-100 for 20 minutes at room temperature. After 1hour blocking in PBS containing 3% bovine serum albu-min cells were incubated with the primary antibody inPBS plus 3% bovine serum albumin, washed thoroughlysix times in PBS, and incubated with the secondary an-

tibody for 1 hour and finally washed six times in PBSagain. Coverslips were mounted using Mowiol 4-88 anti-fade reagent (prepared from polyvinyl alcohol 4-88, glyc-erol, H2O, and Tris pH 8.5).

Confocal images were collected on an LSM510 laserscanning confocal unit (Carl Zeiss) with a 63 � 1.4numerical aperture plan Apochromat and a 40 � 1.3numerical aperture plan Neofluar objective (Carl Zeiss). Ex-citation was with 25-mW argon laser emitting 488 nm,and a 1.0-mW helium/neon laser emitting at 543 nm.Emissions were collected using a 500- to 530-nm bandpass filter to collect A488 and a 560-nm long pass filterto collect A568 emission. Usually images from opticalslices of 1- to 2-�m thickness were acquired. Crosstalk of the fluorophores was negligible.

Figure 1. Characterization of PXDN expression and activity. Detection ofPXDN (A) mRNA expression by Northern blot analysis. Multiple-tissue (2�g of poly�A� � RNA) Northern blot membranes were probed at 65°C witha randomly radiolabeled cDNA fragments corresponding the 3�-untrans-lated regions of PXDN. B: Detection of PXDN by Western blot analysis. Apolyclonal antibody raised against PXDN recognizes the protein in PXDN-expressing COS-7 cells (�, second lane), whereas it does not produceimmunoreactive band in mock-transfected COS-7 cells (�, first lane).Loading controls developed for �-actin indicate that all lanes containedsimilar amounts of total protein. C: Detection of peroxidase activity in thelysates of PXDN-expressing COS-7 cells. COS-7 cells were transfectedwith PXDN cDNA; control cells were mock-transfected. After 48 hourscells were lysed in 1% hexadecyltrimethylammonium bromide and wereassayed for peroxidase activity by the Amplex Red peroxidase assay. Thefluorescent product, resorufin fluorescence was measured at 590 nm.D–F: Intracellular localization of PXDN in transfected COS-7 cells. PXDN-transfected, paraformaldehyde-fixed, permeabilized COS-7 cells werestained for PXDN (D) and the endoplasmic reticulum marker PDI (E).Merge of the fluorescent signals is shown in (F). Scale bar � 20 �m.

Matrix Formation by Myofibroblasts 727AJP August 2009, Vol. 175, No. 2

Gene Expression Studies

For the human PXDN mRNA detection, human multiple-tissue (2 �g of poly�A�� RNA) Northern blot membranes(Clontech) were probed at 65°C with a randomly radio-labeled cDNA fragments corresponding the 3�-untrans-lated regions of PXDN mRNA following standard hybrid-ization methods. For the quantitative PCR experimentsRNA was isolated with Trizol reagent (Invitrogen) accord-ing to the manufacturer’s instructions. cDNA was synthe-sized from 2 �g total RNA using oligo(dt)18 primers andRevertAid M-MuLV Reverse Transcriptase in 20 �l reac-tion mix according to the manufacturer’s (Fermentas)recommendations. One �l of the first strand was ampli-fied in 10 �l total volume in a LightCycler 1.5 instrument(Roche) using LightCycler FastStart DNA Master SYBRGreen I mix (Roche) with a final Mg2� concentration of2.25 mmol/L, and final primer concentration of 0.5 �mol/L.To avoid amplification of the genomic region, primerswere designed in separate exons neighboring long in-trons; the sequences of the primers are provided in Table1. The following PCR protocol was used: 95°C for 10minutes, then 40 cycles of 95°C 10 seconds, 60°C 5seconds, 72°C 15 to 25 seconds (amplicon size �bp�/25),the last step was a melting curve analysis (from 65 to95°C with a slope of 0.1°C/sec). The quantification wasperformed by the LightCycler Software 4.05 as follows.The crossing point was determined by the second deriv-ative method. PCR efficiency and standard curve wascalculated for each gene by performing amplification onserial dilutions of a mixture of the samples. In each sam-ple the expression of the target gene was divided with theexpression of the endogenous control, which was thehousekeeping gene GAPDH. The relative expression lev-

els were finally normalized to the average expressionlevel of the control samples (set to 1).

Animal Model

Animal experiments were authorized by the InstitutionalAnimal Experiment Committee under permission No. 86/2006 SE TUKEB. Eight-week-old male BALB/c mice wereobtained from the National Institute of Oncology. Animalswere maintained on standard diet and given water adlibitum. Unilateral ureteral obstruction was performed us-ing a standard procedure11 Briefly, under ketamine (50mg/kg) and xylazine (10 mg/kg) induced general anes-thesia complete ureteral obstruction was performed byligating the left ureter with 8�0 silk after a midline ab-dominal incision. Mice were sacrificed after 7 days of theprocedure and the kidneys (both obstructed and control)were removed. One part of the kidneys was fixed in 4%phosphate-buffered paraformaldehyde followed by par-affin embedding for histological analysis. Goldner’strichome staining was used for the detection of fibrosis.Histological analysis was performed by HistopathologyLtd, Pecs, Hungary. Samples were coded and examinedin a blinded fashion. The remaining kidneys were pro-cessed for RNA isolation and immunohistochemistry. Weused acetone fixed, 8-�m thick frozen sections for theimmunolocalization of PXDN in control and fibrotic kid-neys. After blocking in 1% albumin and 2% normal goatserum we incubated the sections overnight with anti-PXDN and anti-fibronectin antibodies. Sections were thenstained with fluorophore-labeled secondary antibodiesfor 1 hour at room temperature.

Figure 2. Detection of PXDN protein expression in human pulmonary fibroblasts (HPF) and human dermal fibroblasts (HDF). A: Detection of PXDN byWestern blot analysis in HPFs (first lane) and HDFs (second lane). A polyclonal, PXDN-specific antibody detects PXDN in both cell types. B–G: Intracellularlocalization of PXDN in human primary fibroblasts. Paraformaldehyde-fixed, permeabilized HPFs (B–D) or HDFs (E–G) were stained for PXDN (B, E) orfor the endoplasmic reticulum marker PDI (C, F). Merge of red and green fluorescences is shown in (D) and (G). Note the significant overlap of the twofluorescent signals. Scale bar � 20 �m.


Results

Characterization of the Expression Pattern ofPXDN

Conflicting data have been published regarding the tis-sue expression pattern of PXDN. While Horikoshi et alsuggested that PXDN is ubiquitously expressed in a va-riety of tissues,9 Mitchell et al proposed that it is primarily

a melanoma-specific protein.10 To clarify these conflict-ing data we have used the Northern blot technique tostudy the expression of PXDN mRNA. To exclude thepotential cross-hybridization to mRNAs encoding otherperoxidases, we have used the 3�-untranslated region ofthe PXDN cDNA as a probe. As it is shown in Figure 1A,PXDN mRNA was expressed in several human tissuesincluding heart, skeletal muscle, colon, spleen, kidney,liver, small intestine, and placenta, while it was absent inbrain, thymus, and leukocytes. A similar picture of ex-pression was also suggested by the analysis of PXDNEST sequences deposited in GenBank.

Intracellular Localization and Enzymatic Activityof PXDN

To study the peroxidase activity of PXDN we have ex-pressed the protein in COS-7 cells. Using polyclonalantibodies raised against the protein we could detectPXDN in transfected cells, while mock-transfected cellsshowed no detectable expression (Figure 1B). We usedthe Amplex Red assay to study the peroxidase activity oftransfected cell lysates. Figure 1C shows that we coulddetect peroxidase activity only in PXDN-expressing cells,while lysates of mock-transfected cells showed no de-tectable enzymatic activity. Next we studied the intracel-lular localization of PXDN in transfected COS-7 cells.Figure 1, D–F show that PXDN colocalized with PDI, awell characterized marker protein of the endoplasmicreticulum.12

Next we sought to examine if we can detect endog-enously expressed PXDN in human primary cells. Amongthe cells examined we could detect PXDN protein ex-pression in human pulmonary fibroblasts (HPFs) and hu-man dermal fibroblasts (HDFs) (Figure 2A) and vascularendothelial and smooth muscle cells (data not shown).We used confocal microscopy to study the intracellulardistribution of endogenously expressed PXDN protein. In

Figure 3. TGF-�1 increases PXDN expression in human pulmonary fibro-blasts (HPF). A: Induction of PXDN mRNA expression by TGF-�1 treatment.HPFs were serum-deprived in the presence of 0.05% serum for 24 hours andwere subsequently treated with 5 ng/ml TGF-�1 for 24 hours. RNA wasisolated from the cells and cDNA was synthesized (see Materials and Meth-ods). Quantitative PCR experiments were performed with the SybrGreenmethod. Relative expression levels of PXDN are shown using GAPDH asinternal control. The PXDN expression level in uninduced cells is defined as1. Values are the mean � SEM. B: Induction of PXDN protein expression byTGF-�1 in HPFs. HPFs were serum-deprived in the presence of 0.05% serumfor 24 hours and were subsequently treated with 5 ng/ml TGF-�1 (�) for 24,48, and 72 hours. In the medium of the control cells TGF-�1 was omitted (�).Western blot analysis was used for PXDN detection (upper panel). Detectionof Lamin A and C in loading controls indicate that each lane contained similaramount of protein. C: Detection of TGF-�1-induced PXDN expression byimmunofluorescence. HPFs were serum-deprived in the presence of 0.05%serum for 24 hours and were subsequently treated with 5 ng/ml TGF-�1 for24 hours (D) or left untreated (C). Paraformaldehyde-fixed, permeabilizedcells were stained for PXDN (red color in C and D) and for the myofibroblastmarker SMA (insets in C and D). The appearance of SMA expressionindicates myofibroblastic differentiation. Scale bars � 20 �m. E: Detection ofPXDN in cell culture medium. After 72 hours incubation time TCA was usedto precipitate proteins from the medium of control (untreated) and TGF-�1-treated HDFs. PXDN expression of the cells was analyzed in parallel exper-iments. Loading controls developed for �-actin indicate that cell lysatescontained similar amounts of total protein, while the absence of �-actin in theprecipitated samples suggests that PXDN in the medium does not originatefrom unattached cells.


HPFs and HDFs we observed a reticular staining pattern(Figure 2, B and E). The specificity of staining was con-firmed by two different PXDN-specific siRNAs (data notshown). The observed intracellular localization sug-gested that PXDN localized to the endoplasmic reticulumtherefore we investigated the localization of PXDN inrelation to this organelle. We examined the localizationsof PXDN and PDI. When the cells were stained for PDIand PXDN, we observed significant overlap of the twosignals suggesting that PXDN indeed localized to theendoplasmic reticulum (Figure 2, C–G). Besides its local-ization to the endoplasmic reticulum we have also ob-served perinuclear localization of PXDN in both HPF andHDF cells (Figure 2, D and G).

Induction of PXDN Expression by TGF-�1 inPulmonary and Dermal Fibroblasts

PXDN contains several domains, including leucine-richrepeats, immunoglobulin C2-type domains, which areusually found in proteins of the ECM.13,14 In accordancewith its structural features the Drosophila homolog wasdescribed to be secreted by Kc cells.8 The protein local-ization software TargetP 1.1 predicts that human PXDN,which contains a signal peptide, goes through the secre-tory pathway as well. We therefore sought to find condi-tions when increased protein secretion occurs. Whenprimary fibroblasts are stimulated by TGF-�1 they gothrough a drastic phenotypic change and differentiateinto myofibroblasts. Myofibroblasts possess contractilefeatures and show intense synthesis of ECM proteinsincluding fibronectin and different types of collagen.15,16

We studied the effect of TGF-�1 on PXDN expression inHPFs. As shown in Figure 3A, we have observed a two-fold increase in PXDN mRNA expression after 10 hours

treatment with TGF-�1. Figure 3B shows that after 24hours of TGF treatment we detected an increased level ofPXDN protein by Western blot and elevated PXDN levelwas still observed after 72 hours. We also confirmed thisresult by immunostaining experiments (Figure 3, C andD) where an increase in PXDN protein level was alsoevident after 24 hours and remained elevated for 72hours (data not shown). Induction of the myofibroblastphenotype was confirmed by immunostaining of SMA,which was absent in fibroblasts (see inset in Figure 3C),but appeared during the course of TGF-�1 treatment(inset in Figure 3D). We were interested if PXDN wassecreted to the extracellular space, therefore we ana-lyzed the cell culture medium for its PXDN content. Figure3E shows that TGF-�1 stimulated the secretion of PXDNto the cell culture medium. PXDN in the medium did notoriginate from unattached cells since we could not detect�-actin in the protein precipitate with a highly sensitiveantibody.

Localization of PXDN to Fibril-Like Structures inthe Extracellular Space

During myofibroblast differentiation cells undergo a dras-tic phenotypic change that is accompanied by the in-creased expression of heme-containing proteins such asthe NADPH oxidase Nox417 and PXDN. Since we appliedTGF-�1 to serum-deprived HPFs we have hypothesizedthat the cells may lack enough heme to complete thesynthesis of PXDN. We therefore supplemented the me-dium with �-aminolevulonic acid (ALA), a precursor ofheme biosynthesis.18 Supplementation of the cell culturemedium with ALA during the differentiation process didnot increase the PXDN content of the medium (data notshown), however we observed a drastic change in PXDN

Figure 4. Formation of PXDN-containing fibrils by TGF-�1-treated HPFs and HDFs. A–D: HPFs (A,B) and HDFs (C,D) were serum-deprived in the presence of0.05% serum for 24 hours and were subsequently treated with 5 ng/ml TGF-�1 for 72 hours in the presence of 500 �mol/L ALA. TGF-�1 was omitted from themedium of control cells (A and C). Paraformaldehyde-fixed, permeabilized cells were stained for PXDN. Note the intense staining of fibril-like structures (indicatedby arrows) around the cells (B, D). E. Effect of siRNA treatment on PXDN expression detected by Western blot. HDFs were transfected with PXDNsequence-specific siRNAs or “minimally” changed control siRNAs or left untransfected. Myofibroblast differentiation was induced by 5 ng/ml TGF-�1 for 72 hoursin the presence of 500 �mol/L ALA. Detection �-actin in loading controls indicate that each lane contained similar amount of protein. F–I: Effect of siRNA treatmenton the development of PXDN-containing fibrils. HPFs (F,G) and HDFs (H,I) were transfected with PXDN sequence-specific siRNAs (F,H) or “minimally” changedcontrol siRNA (G,I). siRNAs were transfected to the cells during serum deprivation for 24 hours and myofibroblastic differentiation was subsequently induced by5 ng/ml TGF-�1 in the presence of 500 �mol/L ALA. Paraformaldehyde-fixed, permeabilized cells were stained for PXDN and for the myofibroblast marker SMA.Insets (F–I) show SMA staining from the same microscopic field. Scale bars � 20 �m.


localization (Figure 4). After 72 hours treatment, PXDNappeared in the form of a dense network extracellulardeposition of PXDN (Figure 4, A and B). The appearanceof these structures was also observed in myofibroblastsdeveloped from HDFs, suggesting that the characteristiclocalization of PXDN was not specific for pulmonary cells(Figure 4, C and D). PXDN-specific siRNAs effectivelyinhibited PXDN expression (Figure 4E) and the develop-ment of PXDN-containing fibrils (Figure 4, F–I). Impor-tantly, siRNA treatment did not affect the myofibroblasticdifferentiation, as judged by the expression SMA in thesiRNA-treated cells (insets, Figure 4, F–I). In our nextexperiments we have performed a detailed characteriza-tion of the PXDN-containing structures. Figure 5, A and Bshows that PXDN-containing fibrils are frequently formedbetween neighboring cells and anchor to the proximity ofnucleus, however fibrils bridging larger distances werealso observed. To confirm the extracellular localization ofPXDN we compared the staining pattern of PXDN be-tween nonpermeabilized and permeabilized cells. Asshown in Figure 5C PXDN-containing fibrils were de-tected when cells were not permeabilized. Importantly

the intracellular localization of PXDN was undetectableand the cells did not stain for SMA, confirming the integ-rity of the plasma membrane (Figure 5D). Parallel exper-iment on permeabilized cells proved that SMA indeedappeared in the cells during differentiation (Figure 5, Eand F). Next we sought to determine the relation of PXDN-fibrils to fibronectin. Figure 6 shows that PXDN fibrilspartially colocalize with fibronectin, in cultures of TGF-�1-treated HPFs (Figure 6, A–C) and HDFs (Figure 6, D–F).Colocalization was frequently observed in thick cable-likestructures, while the two proteins did not colocalize inthinner fibrils.

Role of H2O2 and Peroxidase Activity in theSecretion and Localization of PXDN

We analyzed the peroxidase activity of fibroblast anddifferentiated myofibroblast cultures but despite theobvious difference in PXDN expression we could notdetect the peroxidase activity of PXDN (data notshown). Since PXDN acts as a functional peroxidasewhen expressed in COS7 cells (Figure 1C) and TGF-�1-stimulated fibroblasts were described to producereactive oxygen species (ROS)19 we wanted to inves-tigate if we can detect the H2O2-consuming activity ofPXDN. Confirming the results of earlier reports, wemeasured increased H2O2 production by TGF-�1-stim-ulated fibroblasts (Figure 7A). We have hypothesizedthat if PXDN used the produced H2O2, then decreasedPXDN expression should result increased release ofH2O2. As shown in Figure 7A, inhibition of PXDN ex-pression by PXDN-specific siRNA had no effect on theTGF-�1-induced H2O2 production. Next we measuredstimulated H2O2 production at different time pointsafter the TGF-�1 stimulus (Figure 7B). H2O2 productionhas peaked at 24 hours and no stimulated ROS-releasewas observed after 72 hours, when PXDN-containingfibrils appeared in the cell culture.

PXDN Expression Is Increased in a MurineModel of Kidney Fibrosis

We have also studied the possible changes of PXDNexpression in a well-characterized mouse model of kid-ney fibrosis, induced by unilateral ligation of the ureter.We chose this model because the rapid development ofkidney fibrosis in this model is dependent on TGF-�1signaling20 and it has been previously shown that myofi-broblasts appear during the process and secrete largeamount of ECM proteins.20 Figure 8A shows, that after 7days of ligation we have observed a more than threefoldincrease in PXDN expression. Intense blue staining ofkidney tissue with Goldner’s trichome method (Figure 8,B and C) indicated that fibrotic remodeling of the kidneyparalleled the increased expression of PXDN. We couldalso demonstrate the increase of PXDN expression infibrotic kidneys (Figure 8, D–I). While PXDN was barelydetectable in normal kidneys (Figure 8G), the proteinbecomes enriched in the peritubular space of fibrotic

Figure 5. PXDN-containing fibrils are frequently formed between neighbor-ing cells and localize to the extracellular space. A–B: HDFs (A) and HPFs (B)were serum-deprived in the presence of 0.05% serum for 24 hours and weresubsequently treated with 5 ng/ml TGF-�1 for 72 hours in the presence of500 �mol/L ALA. Paraformaldehyde-fixed, permeabilized cells were stainedfor PXDN (A) or PXDN (red) and Lamin (green) (B). White arrows indicatethat PXDN-fibrils frequently dock to the proximity of nuclei. C–F: HDFs wereserum-deprived in the presence of 0.05% serum for 24 hours and weresubsequently treated with 5 ng/ml TGF-�1 for 72 hours in the presence of500 �mol/L ALA. Paraformaldehyde-fixed, non-permeabilized (C,D) or per-meabilized (E,F) cells were stained for PXDN and SMA. Scale bar � 20 �m.


kidneys where it colocalized with fibronectin (Figure 8,D–I). This observation suggests that the stimulatory effectof TGF-�1 on PXDN expression and secretion was notrestricted to in vitro conditions.

Discussion

Tissue fibrosis represents a leading cause of morbidityand mortality worldwide. Fibrosis can arise in nearly ev-ery organ and it is especially important in the develop-ment of kidney diseases, liver cirrhosis, heart disease,and interstitial lung disease.21–23 Emerging evidencesuggests that regardless of the location, myofibroblastsseem to have a central role in the development of fibrosis.Myofibroblasts also have an important physiological rolein wound healing and development. These fascinatingcells can develop from fibroblasts and also from epithelialcells during the course of epithelial–mesenchymal tran-sition. During their differentiation, myofibroblasts developa contractile phenotype due to increased expression ofSMA and they also show increased synthesis of ECMproteins, including fibronectin and different types of col-lagen.15,16 In this work we have found that myofibroblastsalso synthesize and secrete PXDN, a mammalian perox-idase with previously unknown function. Our results sug-gest that PXDN deposited in the extracellular space pro-vides a previously unprecedented pathway of ECMformation.

PXDN is a unique member of the peroxidase familybecause its catalytic domain is surrounded by domainscharacteristic for proteins of the ECM. These parts in-clude leucine-rich repeats and C2-type immunoglobulindomains, which localize to the N-terminus of the protein,while a vWF C-type domain is localized to the C-terminalpart of the protein. PXDN was first described in Drosoph-ila melanogaster.7 Kc cells of Drosophila origin secreteperoxidasin to the extracellular medium, where the pro-tein exists in trimers. Based on the expression pattern ofits mRNA, Drosophila PXDN is present in hemocytes andthought to be involved in ECM synthesis during fly devel-opment. There is scant amount of information about thehuman PXDN. Horikoshi et al isolated cDNAs from cellswhere apoptosis was induced in a p53-dependent way.9

Among the induced cDNAs, they found an alternatively

Figure 6. PXDN partially colocalizes with fi-bronectin in PXDN-contaning fibrils. HPFs(A–C) and HDFs (D–F) were serum-deprived inthe presence of 0.05% serum for 24 hours andwere subsequently treated with 5 ng/ml TGF-�1for 72 hours in the presence of 500 �mol/L ALA.Paraformaldehyde-fixed, permeabilized cellswere stained for PXDN (A,D) and fibronectin(B,E). Note the intense staining of fibril-likestructures around the cells, whereas the intracel-lular staining is essentially absent. Merge of redand green fluorescences are shown in (C) and(F). Note the colocalization of PXDN and fi-bronectin in large, cable-like structures. Scalebar � 20 �m.

PXDN-1 siRNA control siRNA0

0,5

1

1,5

2

2,5

H2O

2 pro

duct

ion

(fold

incr

ease

)

A

0,8

1

1,2

1,4

1,6

1,8

2

H2O

2 pro

duct

ion

(fold

incr

ease

)

B

12 h 24 h 36 h 48 h

Figure 7. Role of hydrogen peroxide and peroxidase activity in PXDNfunction. A: Effect of PXDN-specific siRNA treatment on TGF-�1-inducedH2O2 production. HPFs were transfected with PXDN sequence-specific siR-NAs or “minimally” changed control siRNS. siRNAs were transfected to thecells during serum deprivation for 24 hours and cells were treated with 5ng/ml TGF-�1 in the presence of 500 �mol/L ALA for 24 hours. H2O2

production was measured by the Amplex Red assay. H2O2 production ofuninduced cells is defined as 1. B: Kinetics of TGF-�1 induced H2O2 pro-duction. HPFs were serum-deprived in the presence of 0.05% serum for 24hours and were subsequently treated with 5 ng/ml TGF-�1 for the indicatedperiods of time. H2O2 production of uninduced cells is defined as 1.


spliced product of the human PXDN gene. The authorshave hypothesized that the induction of PXDN might becoupled to the altered ROS metabolism in apoptosis, butthey did not elaborate on this idea. According to theirstudy PXDN is expressed in almost all human tissues. Inan independent study PXDN was identified as a novelmelanoma gene (MG50) based on its expression in mel-anoma samples and relative absence in normal tissuesamples.10 Our results are in agreement with the report ofHorikoshi et al since we have detected PXDN mRNAexpression in several tissues, with particularly high ex-pression level in heart, skeletal muscle, small intestineand placenta (Figure 1).9 PXDN functions as a peroxidasewhen expressed in COS-7 cells (Figure 1). To further char-acterize the peroxidase activity of a purified protein we havealso expressed an epitope-tagged form of PXDN in Sf9insect cells and attempted to purify it. Unfortunately, theprotein remained in the insoluble fraction during purificationand we have been unable to recover functional protein fromthat fraction (data not shown).

In our experiments TGF-�1 increased the expressionof PXDN and its secretion to the cell culture medium(Figure 3). Myofibroblastic differentiation is usually in-duced in serum-deprived cells, since responsiveness toTGF-�1 is increased under those conditions. We havehypothesized that in the absence of serum, the cellularproduction of heme might not be sufficient for the syn-thesis of heme-containing peroxidases. This assumptionwas also originated from the observation that TGF-�1

increases the expression of another heme-containingprotein, the NADPH oxidase Nox4.17 In our experimentswe observed a more than 100-fold increase in Nox4expression at mRNA level (data not shown) proving amarkedly increased demand for heme, during the myofi-broblastic differentiation. When we have supplementedthe medium with ALA, a precursor of heme biosynthesis,we could observe the localization of PXDN into fibril-likestructures (Figure 4). Although heterologously expressedPXDN exhibited peroxidase activity we could not detectthe enzymatic activity of PXDN in fibroblasts and differ-entiated myofibroblasts. The lack of measurable peroxi-dase activity might be explained by the fact that PXDNhas very low enzymatic activity when compared withother peroxidases. For example COS7 cells, that heter-ologously express lactoperoxidase, show 100-fold higherperoxidase activity than PXDN-expressing COS7 cells(data not shown). Nelson et al has also shown, that com-pared with other peroxidases, Drosophila PXDN hadlower affinity for classical peroxidase substrates.7 Alter-natively, it is also possible that endogenously expressedPXDN is in complex with another protein that inhibits theenzymatic activity of PXDN. Incorporation of PXDN intolarger complexes was also suggested by Nelson et alwho could not recover active PXDN from Drosophila tis-sues showing high level expression of the protein.7 Nev-ertheless we cannot exclude the possibility that PXDN stillfunctions as a peroxidase in the endoplasmic reticulumor in the extracellular matrix of myofibroblasts. Since we

Figure 8. PXDN expression is increased in aTGF-�1-dependent model of kidney fibrosis. A:Induction of PXDN mRNA expression by unilat-eral ureteral obstruction. Unilateral ureteral ob-struction was applied for 7 days after that kid-neys were removed and their PXDN expressionwas measured by quantitative PCR analysis per-formed with the SybrGreen method. Relativeexpression levels of PXDN are shown usingGAPDH as internal control. The PXDN expres-sion level in control, non-obstructed kidneys isdefined as 1. Values are the mean � SEM. B, C:Detection of unilateral ureteral obstruction-in-duced kidney fibrosis by Goldner trichromestaining. After 7 days of unilateral ureteral ob-struction, kidneys were removed and fixed in4% phosphate-buffered paraformaldehyde. Af-ter paraffin embedding, sections were made andstained by Goldner’s trichrome method. Blue-staining, marked by black arrows indicates thedevelopment of fibrosis in the peritubular spaceof obstructed kidneys (B). Untreated kidneysfrom the same animals served as controls (C).D–I: Detection of PXDN and fibronectin in con-trol and obstructed kidneys. Acetone fixed, fro-zen sections were stained for PXDN (D,G) andfibronectin (E,H). Merge of red and green fluo-rescences are shown in (F) and (I). Note theperitubular co-localization of PXDN and fi-bronectin. Scale bar � 20 �m.


could not measure the enzymatic activity of PXDN inmyofibroblast cultures we cannot state that ALA indeedincreased the maturation of PXDN. However without ALAsupplementation TGF-�1 did not induce the appearanceof PXDN in extracellular, fibril-like structures. Since inclu-sion of ALA during myofibroblast differentiation did notchange the amount of precipitable PXDN in the medium,it is possible that formation of PXDN-containing fibrils is aresult of a different secretory pathway that is uncoveredby assisted heme synthesis. Importantly in vivo, duringkidney fibrosis PXDN appears in the peritubular spaceproviding further evidence for its secretion (Figure 8). Thepresence of PXDN-containing fibrils in the extracellularspace, suggested the multimerization of the secretedprotein and its association to other constituents of theECM. In fact we could detect partial colocalization ofPXDN and fibronectin around the cells (Figure 6). Colo-calization of PXDN and fibronectin was also observed infibrotic kidneys. The domain organization of PXDN sug-gests that the protein is ideally suited for the stabilizationof the ECM through protein–protein interactions. Dro-sophila PXDN was described to exist in a trimeric form, acomplex that is thought to be organized through its C-terminal vWF C-type domain.7 The same domain is alsorecognized in mammalian PXDN. This domain is alsofound in other ECM proteins including the throm-bospondin type I and II group of vertebrate matrix pro-teins, which also exist in the form of trimers.24 The pres-ence of leucine-rich repeats and immunoglobulin loops inPXDN also suggests that this protein readily associateswith other ECM proteins. Leucine rich repeats are shortsequence motifs present in a number of proteins withdiverse functions.13 These protein modules are usuallyinvolved in homophilic and heterophilic protein-proteininteractions. Immunoglobulin loop motifs are also fre-quently found in neural cell adhesion molecules such asfascilin and neuroglian.14 The combination leucine-richrepeats and immunoglobulin C2 type domains is notfrequently observed in proteins. In fact the only otherknown mammalian example for such combination is theLRIG (leucine-rich repeats and immunoglobulin-like do-mains) family of proteins, members of which are trans-membrane proteins and were previously shown to neg-atively regulate the ErbB family of receptor tyrosinekinases.25,26

Peroxidases in lower species are involved in stabiliza-tion of the ECM through tyrosine–tyrosine crosslinks.5

Oxidative formation of dityrosines contributes to the cuti-cle synthesis in C. elegans and hardening of the fertiliza-tion envelope in sea urchin eggs.7,27 The ability to formdityrosine crosslinks is a general feature of mammalianperoxidases as well, but the physiological significance ofsuch activity remains to be established. It was reason-able to assume that crosslinking peroxidase activity wasinvolved in the formation of PXDN-containing fibrils, sinceThannickal et al described increased H2O2 production byTGF-�1-stimulated lung fibroblasts.19,28 In subsequentstudies they have also found that H2O2 produced byfibroblasts indeed supports dityrosine formation by anexogenous peroxidase.28 In our experiments however, itseems unlikely, since we could not detect the peroxidase

activity of PXDN in myofibroblasts and no increase indityrosine level was observed during the course of myo-fibroblast differentiation (data not shown). On the otherhand, since the peroxidase domain of PXDN is intact wewere interested if reactive oxygen species, more specif-ically H2O2 has any role in the formation of PXDN-con-taining extracellular structures. Our results, however,suggest that TGF-�1-stimulated H2O2 production unlikelyhas a role in the formation of PXDN-containing fibrils.First, suppression of PXDN expression did not affect ROSrelease from the cells, suggesting that H2O2 was notused in peroxidase-catalyzed reaction. Second, the ki-netics of TGF-�1-stimulated H2O2 production did not par-allel the formation of PXDN-containing fibrils, since max-imal H2O2 production was observed at 24 hours afterTGF-�1-stimulation, when fibroblasts have not differenti-ated yet into myofibroblasts. The kinetics of TGF-�1-stimulated ROS production rather suggests that H2O2

has a role in the differentiation process itself. Experimentsby Cucoranu et al suggest that induction of the ROS-producing enzyme Nox4 would be responsible for oxida-tive changes during myofibroblastic differentiation.17

At least three members of the mammalian peroxidasefamily have host defense function, therefore it will beinteresting to explore if secreted PXDN has any antimi-crobial activity. This possibility would certainly be impor-tant during wound healing when myofibroblasts are nor-mally activated. Activated neutrophil granulocytes havebeen recently described to form fibril-like structures,neutrophil extracellular traps, which contained elas-tase, DNA, myeloperoxidase, and showed antimicro-bial activity.29

In summary, we conclude that the synthesis and se-cretion of PXDN by myofibroblasts is a previously unrec-ognized pathway in the formation of the ECM. Futurestudies should identify the function of PXDN in myofibro-blasts and explore the importance of this pathway inwound healing and in the development of diseases wherefibrogenic response seems to be an important part of theunderlying pathology.

Acknowledgments

We are grateful to Andras Kapus and Laszlo Buday forhelpful comments about the manuscript.

References

1. O’Brien PJ: Peroxidases. Chem Biol Interact 2000, 129:113–1392. Klebanoff SJ: Myeloperoxidase: friend and foe. J Leukoc Biol 2005,

77:598–6253. Wijkstrom-Frei C, El Chemaly S, Ali-Rachedi R, Gerson C, Cobas MA,

Forteza R, Salathe M, Conner GE: Lactoperoxidase and human air-way host defense. Am J Respir Cell Mol Biol 2003, 29:206–212

4. Ruf J, Carayon P: Structural and functional aspects of thyroid perox-idase. Arch Biochem Biophys 2006, 445:269–277

5. Daiyasu H, Toh H: Molecular evolution of the myeloperoxidase family.J Mol Evol 2000, 51:433–445

6. LaFleur GJ, Jr., Horiuchi Y, Wessel GM: Sea urchin ovoperoxidase:oocyte-specific member of a heme-dependent peroxidase superfam-ily that functions in the block to polyspermy. Mech Dev 1998,70:77–89


7. Edens WA, Sharling L, Cheng G, Shapira R, Kinkade JM, Lee T,Edens HA, Tang X, Sullards C, Flaherty DB, Benian GM, Lambeth JD:Tyrosine cross-linking of extracellular matrix is catalyzed by Duox, amultidomain oxidase/peroxidase with homology to the phagocyteoxidase subunit gp91phox. J Cell Biol 2001, 154:879–891

8. Nelson RE, Fessler LI, Takagi Y, Blumberg B, Keene DR, Olson PF,Parker CG, Fessler JH: Peroxidasin: a novel enzyme-matrix protein ofDrosophila development. EMBO J 1994, 13:3438–3447

9. Horikoshi N, Cong J, Kley N, Shenk T: Isolation of differentially ex-pressed cDNAs from p53-dependent apoptotic cells: activation of thehuman homologue of the Drosophila peroxidasin gene. BiochemBiophys Res Commun 1999, 261:864–869

10. Mitchell MS, Kan-Mitchell J, Minev B, Edman C, Deans RJ: A novelmelanoma gene (MG50) encoding the interleukin 1 receptor antago-nist and six epitopes recognized by human cytolytic T lymphocytes.Cancer Res 2000, 60:6448–6456

11. Chevalier RL, Thornhill BA, Wolstenholme JT, Kim A: Unilateral ure-teral obstruction in early development alters renal growth: depen-dence on the duration of obstruction. J Urol 1999, 161:309–313

12. Ellgaard L, Ruddock LW: The human protein disulphide isomerasefamily: substrate interactions and functional properties. EMBO Rep2005, 6:28–32

13. Kobe B, Kajava AV: The leucine-rich repeat as a protein recognitionmotif. Curr Opin Struct Biol 2001, 11:725–732

14. Walsh FS, Doherty P: Neural cell adhesion molecules of the immuno-globulin superfamily: role in axon growth and guidance. Annu RevCell Dev Biol 1997, 13:425–456

15. Tomasek JJ, Gabbiani G, Hinz B, Chaponnier C, Brown RA: Myofi-broblasts and mechano-regulation of connective tissue remodelling.Nat Rev Mol Cell Biol 2002, 3:349–363

16. Hinz B, Phan SH, Thannickal VJ, Galli A, Bochaton-Piallat ML, GabbianiG: The myofibroblast. One function, multiple origins. Am J Pathol 2007,170:1807–1816

17. Cucoranu I, Clempus R, Dikalova A, Phelan PJ, Ariyan S, Dikalov S,Sorescu D: NAD(P)H oxidase 4 mediates transforming growth factor-

beta1-induced differentiation of cardiac fibroblasts into myofibro-blasts. Circ Res 2005, 97:900–907

18. Ponka P: Cell biology of heme. Am J Med Sci 1999, 318:241–25619. Thannickal VJ, Fanburg BL: Activation of an H2O2-generating NADH

oxidase in human lung fibroblasts by transforming growth factor beta1. J Biol Chem 1995, 270:30334–30338

20. Yang J, Liu Y: Dissection of key events in tubular epithelial to myofi-broblast transition and its implications in renal interstitial fibrosis.Am J Pathol 2001, 159:1465–1475

21. Wynn TA: Common and unique mechanisms regulate fibrosis invarious fibroproliferative diseases. J Clin Invest 2007, 117:524–529

22. Thannickal VJ, Toews GB, White ES, Lynch JP, III, Martinez FJ:Mechanisms of pulmonary fibrosis. Annu Rev Med 2004, 55:395–417

23. Bataller R, Brenner DA: Liver fibrosis. J Clin Invest 2005, 115:209 –218

24. Bornstein P: Thrombospondins: structure and regulation of expres-sion. FASEB J 1992, 6:3290–3299

25. Suzuki Y, Sato N, Tohyama M, Wanaka A, Takagi T: cDNA cloning of anovel membrane glycoprotein that is expressed specifically in glial cellsin the mouse brain. LIG-1, a protein with leucine-rich repeats and immu-noglobulin-like domains. J Biol Chem 1996, 271:22522–22527

26. Shattuck DL, Miller JK, Laederich M, Funes M, Petersen H, CarrawayKL, III, Sweeney C: LRIG1 is a novel negative regulator of the Metreceptor and opposes Met and Her2 synergy. Mol Cell Biol 2007,27:1934–1946

27. Heinecke JW, Shapiro BM: The respiratory burst oxidase of fertiliza-tion. A physiological target for regulation by protein kinase C. J BiolChem 1992, 267:7959–7962

28. Larios JM, Budhiraja R, Fanburg BL, Thannickal VJ: Oxidative proteincross-linking reactions involving L-tyrosine in transforming growth factor-beta1-stimulated fibroblasts. J Biol Chem 2001, 276:17437–17441

29. Brinkmann V, Reichard U, Goosmann C, Fauler B, Uhlemann Y, WeissDS, Weinrauch Y, Zychlinsky A: Neutrophil extracellular traps killbacteria. Science 2004, 303:1532–1535


Peroxidasin Is Secreted and Incorporated into the Extracellular Matrix of Myofibroblasts and Fibrotic Kidney

Documents